Gait analysis is a valuable approach for understanding human movement, but the space and setup requirements of traditional marker-based systems can limit their use outside specialized laboratories. Markerless motion capture may provide a more flexible option, though its agreement in constrained environments compared with traditional spaces is not well established. This study compared a 10-camera markerless system deployed in a hallway with a traditional 8-camera laboratory setup. Twenty-five healthy adults (15 females, 10 males; age 34 [16] y) completed quiet standing, 60 seconds of self-selected walking, and 5-repetition sit-to-stand tasks at both sites on the same day. Three-dimensional pose estimates were processed to calculate alignment during standing, lower-limb joint kinematics during walking, and trunk flexion during sit-to-stand. Agreement within and between sites was assessed using Pearson correlations, root mean square error, Bland Altman limits of agreement, and intraclass correlation coefficients. Standing and walking outcomes showed excellent agreement (intraclass correlation coefficients >= .97; root mean square error < 2.3 degrees; mean differences < 1.1 degrees). Sit-to-stand was more variable (limits of agreement 12 degrees-20 degrees) but remained highly reliable (intraclass correlation coefficients > .96). These findings indicate that a constrained markerless setup can yield kinematic data comparable to a laboratory arrangement, suggesting potential for broader use of markerless approaches in diverse environments.

Benign bone tumors such as chondroblastoma, giant cell tumors (GCT), and aneurysmal bone cysts (ABC) are rare but clinically significant lesions that frequently occur in the epiphyseal regions of long bones, particularly near load-bearing joints in children and young adults. These tumors compromise the structural integrity of bone, leading to an elevated risk of pathologic fracture. Traditional methods for estimating fracture risk rely on simple geometric thresholds and volumetric ratios, but they fail to account for patient-specific differences in bone geometry, material heterogeneity, and physiological loading conditions. As a result, risk is often misclassified, which may lead to either overtreatment or missed prevention opportunities. To address this limitation, this study presents a preliminary demonstration of computed tomography-based finite element analysis (CTFEA) as a novel alternative method (NAM); computational framework using patient-specific CTFEA to evaluate fracture risk in four patients with benign knee tumors. Clinical computed tomography (CT) imaging and motion capture-informed joint loading were used to develop anatomically accurate, mechanically calibrated models incorporating nonlinear bone behavior. CTFEA simulations focused on walking, jogging, and partial weight-bearing conditions, captured localized stress and strain distributions, and were benchmarked against clinical and volumetric assessment criteria. CTFEA outperformed traditional methods by revealing mechanical vulnerabilities, including in cases classified as low-risk clinically, through its ability to simulate individualized loading scenarios. These findings demonstrate the feasibility and potential of CTFEA as a noninvasive, patient-specific alternative to animal or oversimplified models, with direct implications for preoperative planning and fracture risk stratification in orthopedic surgery.

Markerless motion capture may improve the accessibility and participants experience of three-dimensional gait analysis in children. This study explores aspects of usability of markerless motion capture that have not been previously studied specifically, the consequences of street clothing on lower-limb kinematics, participant and caregiver perceptions, and assessment durations. Thirty typically developing children completed two 3D gait analysis protocols. They wore either a) "Conventional" clothing required for marker-based methods and had markers placed, or b) their "Street" clothes without markers. Markerless gait kinematics were measured using Theia3D. Root-mean-square-deviations (RMSD) and an outlier analysis were used to determine differences between experimental conditions. Differences in participant perceptions were assessed using custom surveys, and differences in testing duration were assessed using the Wilcoxon Signed Rank test. Median RMSDs were < 4° and maximal RMSDs were 3.6°-16.0° and with no consistent pattern across joints and planes of motion, suggesting minimal differences between clothing conditions. Individuals with larger deviations generally wore baggy or loose clothing. When asked which condition they would prefer to repeat, 12% (n = 3/25) indicated the Conventional condition, 68% (n = 17/25) mentioned the Street condition and 16% (n = 4/25) would repeat both. Further, caregivers more frequently reported atypical gait in the Conventional condition. The Street condition took less time with a median (25th-75th percentile) difference of 11 (9-13) minutes (p < 0.001). Street clothing with minimal restrictions can be used without compromising lower-limb kinematics, while improving participant experience and reducing assessment duration. These findings may contribute to improved access to 3D gait analysis.

Markerless motion capture addresses key barriers limiting the clinical uptake of biomechanical assessments by enabling efficient data collection and standardized modeling, making it well-suited for multicentre research. This study assessed whether gait deviations associated with knee osteoarthritis (OA) could be consistently detected using markerless motion capture across three clinical centres in Canada. Gait data from 486 participants (351 with knee OA; 135 controls) were analyzed, with body segment kinematics estimated from video using Theia3D. Principal component analysis and linear models were used to evaluate joint kinematics and temporal-distance parameters across groups and sites. After pooling data across centres, individuals with knee OA exhibited characteristic gait deviations, including slower walking speed, reduced hip, knee, and ankle range of motion, and increased knee adduction, compared to controls. These deviations were observed consistently across all three centres. Inter-site differences in joint kinematics were minor (RMS < 3°), remained within reported inter-site error thresholds from marker-based systems, and did not obscure group-level effects. These findings demonstrate that clinically meaningful gait deviations can be reliably detected using markerless motion capture in varied clinical environments without extensive standardization. This work supports its use in multicentre studies and highlights its potential to enable large-scale biomechanical research, an essential step toward broader clinical integration of movement analysis.

Lower limb alignment and specifically the hip-knee-ankle-angle (HKAA), is increasingly being recognized as an important factor in operative planning for total knee arthroplasty (TKA) for patients with knee osteoarthritis (OA). Current clinical practice is to measure alignment from weight bearing radiographs. This static measurement may differ from alignment during the weightbearing stance phase of gait. Markerless motion capture is a novel biomechanical assessment tool that uses off-the-shelf video cameras to capture whole-body movements. The objective of our study was to evaluate HKAA during static and dynamic tasks for patients with predominantly medial or lateral knee OA. Patients diagnosed with knee OA were recruited from advanced care physiotherapists at an outpatient hospital. Markerless motion capture was used to measure whole body movements during a quiet standing task and during gait using eight synchronized Sony RX0 II video cameras at 60 Hz (Sony, Minato, Japan). During quiet stand, subjects stood upright with their feet facing forward and shoulder-width apart for thirty seconds. Gait was assessed during overground walking at a self-selected speed for one minute. Video data was processed with Theia3D (v. 2023.01.0.361 p7, Theia Markerless Inc. Kingston, ON). Visual3D (C-Motion, Germantown, MD) was used to provide gait event detection and estimate hip, knee, and ankle joint locations in three-dimensions at each frame. HKAA was calculated from the three-dimensional joint locations in the frontal plane. For the static quiet standing task, HKAA was averaged over the first five seconds of the standing task. During gait, dynamic HKAA was calculated as the peak angle prior to terminal stance. Correlation between static and dynamic HKAA were calculated and compared using the Bland and Altman approach. Classification of medial or lateral knee OA was done by an orthopaedic surgeon via radiographic evaluation. Dynamic HKAA was compared between medial and lateral knee OA patients (t-test). Static and dynamic HKAA were assessed for 160 knees in 102 patients (34 male, 68 female, mean age 67 years (SD 9)). The static HKAA from the quiet standing task (mean = 182.7 degrees, SD = 4.2) and the peak dynamic HKAA from the gait task (mean = 184.7 degrees SD = 4.6) were highly correlated (Pearson's r = 0.88, p < 0.001). Dynamic HKAA was on average two degrees more varus . Dynamic HKAA was significantly more varus for patients with medial OA (mean 185.9 degrees) than those with lateral OA (mean 176.7 degrees, p < 0.001, t-test). Dynamic peak HKAA measured with markerless motion capture was found to be strongly correlated with static HKAA. Markerless motion capture provides an alternative, low burden, radiation free assessment tool with high potential for clinical integration. The ability to assess dynamic tasks and dynamic parameters, such as HKAA throughout gait, means that markerless motion capture has the potential to provide enhanced assessments pre- and post-TKA.

PURPOSE:Marked lower limb malalignment is associated with the progression of knee osteoarthritis (OA). Markerless motion capture is a computer-vision tool that can measure lower limb alignment throughout both static and dynamic tasks. RESEARCH QUESTIONS:The primary objective of the study was to quantify static and dynamic lower limb alignment for patients diagnosed with advanced medial and lateral knee OA. This analysis also explored sex differences in alignment from the static and dynamic tasks. The secondary objective was to investigate if the mean knee adduction angle during quiet standing and the peak knee adduction angle during the first half of stance of gait were associated. METHODS:Ninety-two patients (37 male, 55 female) diagnosed with advanced knee OA (83 with predominantly medial knee OA, and 9 with predominantly lateral knee OA) completed a quiet standing task and a gait task for markerless motion capture using Theia3D software. Two-way analysis of variance tests were used to investigate sex differences and unpaired t-tests were computed to compare knee OA groups. RESULTS:Statistically significant differences in the knee adduction angle were found between medial and lateral knee OA groups during quiet standing (p < 0.001) and the first half of stance of gait (p < 0.0001). A strong correlation was found (Spearman's ρ = 0.86, p < 0.0001) in lower limb alignment between the static and gait tasks. SIGNIFICANCE:These results show how markerless technology was able to quantify lower limb alignment and demonstrates potential for integration into clinic to assess musculoskeletal diseases.

Background:Bone stress injury (BSI) is a common overuse injury in female athletes that can occur in a variety of bones, including both proximal (pelvis, sacrum, femoral neck) or distal (tibia, fibula, metatarsals) locations. Prior work has demonstrated differences in running biomechanics in those with BSI; however, this was not separated by anatomy. We hypothesised that both female athletes with distal BSI and female athletes with proximal BSI would have lower cadence, higher centre of mass (COM) and lower duty factor than those without prior BSI. Methods:Cross-sectional study of 45 female athletes (15 with prior distal BSI, 15 with prior proximal BSI and 15 with no BSI history). Each ran on an instrumented treadmill at self-selected and 5-kilometre race speeds, with data collected in a fresh and exerted state. A series of analysis of variance tests (ANOVAs, group by condition) were used to analyse the results. Results:Participants with previous proximal BSI ran with greater vertical COM excursion compared with those with no previous BSI at race speed (10.2±1.7 cm vs 8.5±0.8 cm (p<0.001)). The proximal BSI population had a lower cadence than the no prior BSI population at race speed (170±13 steps per minute vs 180±10 steps per minute (p=0.012)). Duty factor was lower in the proximal BSI group compared with the distal BSI group at the race speed (32±3% vs 34±3% (p=0.013)). Conclusion:COM and cadence should be further investigated for association with proximal BSI.

Recent reports have suggested that there may be a relationship between footstrike pattern and overuse injury incidence and type. With the recent increase in wearable sensors, it is important to identify paradigms where the footstrike pattern can be detected in real-time from minimal data. Machine learning was used to classify tibial acceleration data into three distinct footstrike patterns: rearfoot, midfoot, or forefoot. Tibial accelerometry data were collected during treadmill running from 58 participants who each ran with rearfoot, midfoot, and forefoot strike patterns. These data were used as inputs into an artificial neural network classifier. Models were created by using three distinct acceleration data sets, using the first 100%, 75%, and 40% of stance phase. All models were able to predict the footstrike pattern with up to 89.9% average accuracy. The highest error was associated with the identification of the midfoot versus forefoot strike pattern. This technique required no pre-selection of features or filtering of the data and may be easily incorporated into a wearable device to aid with real-time footstrike pattern detection.

Download This Paper Open PDF in Browser Add Paper to My Library Share: Permalink Using these links will ensure access to this page indefinitely Copy URL Copy DOI

The purpose of this study was to develop a machine learning model to reconstruct time series kinematic and kinetic profiles of the ankle and knee joint across six different tasks using an ankle-mounted IMU. Four male collegiate basketball players performed repeated tasks, including walking, jogging, running, sidestep cutting, max-height jumping, and stop-jumping, resulting in a total of 102 movements. Ankle and knee flexion-extension angles and moments were estimated using motion capture and inverse dynamics and considered 'actual data' for the purpose of model fitting. Synchronous acceleration and angular velocity data were collected from right ankle-mounted IMUs. A time-series feature extraction model was used to determine a set of features used as input to a random forest regression model to predict the ankle and knee kinematics and kinetics. Five-fold cross-validation was performed to verify the model accuracy, and statistical parametric mapping was used to determine the difference between the predicted and experimental time series. The random forest regression model predicted the time-series profiles of the ankle and knee flexion-extension angles and moments with high accuracy (Kinematics: R2 ranged from 0.782 to 0.962, RMSE ranged from 2.19° to 11.58°; Kinetics: R2 ranged from 0.711 to 0.966, RMSE ranged from 0.10 Nm/kg to 0.41 Nm/kg). There were differences between predicted and actual time series for the knee flexion-extension moment during stop-jumping and walking. An appropriately trained feature-based regression model can predict time series knee and ankle joint angles and moments across a wide range of tasks using a single ankle-mounted IMU.

As markerless motion capture is increasingly used to measure 3-dimensional human pose, it is important to understand how markerless results can be interpreted alongside historical marker-based data and how they are impacted by clothing. We compared concurrent running kinematics and kinetics between marker-based and markerless motion capture, and between 2 markerless clothing conditions. Thirty adults ran on an instrumented treadmill wearing motion capture clothing while concurrent marker-based and markerless data were recorded, and ran a second time wearing athletic clothing (shorts and t-shirt) while markerless data were recorded. Differences calculated between the concurrent signals from both systems, and also between each participant’s mean signals from both asynchronous clothing conditions were summarized across all participants using root mean square differences. Most kinematic and kinetic signals were visually consistent between systems and markerless clothing conditions. Between systems, joint center positions differed by 3 cm or less, sagittal plane joint angles differed by 5° or less, and frontal and transverse plane angles differed by 5° to 10°. Joint moments differed by 0.3 N·m/kg or less between systems. Differences were sensitive to segment coordinate system definitions, highlighting the effects of these definitions when comparing against historical data or other motion capture modalities.

Motion analysis has seen minimal adoption for orthopaedic clinical assessments. Markerless motion capture solutions, namely Theia3D, address limitations of previous methods and provide gait outcomes that are robust to clothing choice and repeatable in healthy adults. Repeatability in orthopaedic populations has not been investigated and is important for clinical utility and adoption. The purpose of this study was to evaluate the repeatability of Theia3D for gait analysis in a knee osteoarthritis population. Ten orthopaedic patients with knee osteoarthritis underwent gait analysis on three visits, with an average of 8 days between. Participants were recorded during one-minute overground walking trials at self-selected typical and fast speeds by 8 synchronized video cameras. Video data were processed using Theia3D. Intraclass correlations were used to examine the repeatability of temporal distance metrics as well as segment lengths of the underlying kinematic model. Inter-trial and inter-session variability of lower extremity joint angles were estimated for each point of the gait cycle. Intraclass correlations were greater than 0.98 for all temporal distance metrics for both speeds. Lower body segment lengths had intraclass correlations above 0.90. Participant average joint angle waveforms displayed consistent patterns between visits. The average inter-trial and inter-session variability in joint angles across speeds were 1.17 and 1.45 degrees, respectively. The variability in joint angles between visits was less than typically reported for marker-based methods. Gait outcomes measured with Theia3D were highly repeatable in patients with knee osteoarthritis providing further validation for its use in clinical assessment and longitudinal studies.

Background: Bone stress injury (BSI) is a common overuse injury in active women. BSIs can be classified as high-risk (pelvis, sacrum, and femoral neck) or low-risk (tibia, fibula, and metatarsals). Risk factors for BSI include low energy availability, menstrual dysfunction, and poor bone health. Higher vertical load rates during running have been observed in women with a history of BSI. Purpose/Hypothesis: The purpose of this study was to characterize factors associated with BSI in a population of premenopausal women, comparing those with a history of high-risk or low-risk BSI with those with no history of BSI. It was hypothesized that women with a history of high-risk BSI would be more likely to exhibit lower bone mineral density (BMD) and related factors and less favorable bone microarchitecture compared with women with a history of low-risk BSI. In contrast, women with a history of low-risk BSI would have higher load rates. Study Design: Cross-sectional study; Level of evidence, 3. Methods: Enrolled were 15 women with a history of high-risk BSI, 15 with a history of low-risk BSI, and 15 with no history of BSI. BMD for the whole body, hip, and spine was standardized using z scores on dual-energy x-ray absorptiometry. High-resolution peripheral quantitative computed tomography was used to quantify bone microarchitecture at the radius and distal tibia. Participants completed surveys characterizing factors that influence bone health—including sleep, menstrual history, and eating behaviors—utilizing the Eating Disorder Examination Questionnaire (EDE-Q). Each participant completed a biomechanical assessment using an instrumented treadmill to measure load rates before and after a run to exertion. Results: Women with a history of high-risk BSI had lower spine z scores than those with low-risk BSI (–1.04 ± 0.76 vs –0.01 ± 1.15; P < .05). Women with a history of high-risk BSI, compared with low-risk BSI and no BSI, had the highest EDE-Q subscores for Shape Concern (1.46 ± 1.28 vs 0.76 ± 0.78 and 0.43 ± 0.43) and Eating Concern (0.55 ± 0.75 vs 0.16 ± 0.38 and 0.11 ± 0.21), as well as the greatest difference between minimum and maximum weight at current height (11.3 ± 5.4 vs 7.7 ± 2.9 and 7.6 ± 3.3 kg) ( P < .05 for all). Women with a history of high-risk BSI were more likely than those with no history of BSI to sleep <7 hours on average per night during the week (80% vs 33.3%; P < .05). The mean and instantaneous vertical load rates were not different between groups. Conclusion: Women with a history of high-risk BSI were more likely to exhibit risk factors for poor bone health, including lower BMD, while load rates did not distinguish women with a history of BSI.

Background: Bone stress injuries (BSIs) are common in female runners, and recurrent BSI rates are high. Previous work suggests an association between higher impact loading during running and tibial BSI. However, it is unknown whether impact loading and fatigue-related loading changes discriminate women with a history of multiple BSIs. This study compared impact variables at the beginning of a treadmill run to exertion and the changes in those variables with exertion among female runners with no history of BSI as well as among those with a history of single or multiple BSIs. Methods: We enrolled 45 female runners (aged 18-40 years) for this cross-sectional study: having no history of diagnosed lower extremity BSI (N-BSI, n = 14); a history of 1 lower extremity BSI (1-BSI, n = 16); and diagnosed by imaging, or a history of multiple (>= 3) lower extremity BSIs (M-BSI, n = 15). Participants completed a 5-km race speed run on an instrumented treadmill while wearing an Inertial Measurement Unit. The vertical average loading rate (VALR), vertical instantaneous loading rate (VILR), vertical stiffness during impact via instrumented treadmill, and tibial shock determined as the peak positive tibial acceleration via Inertial Measurement Unit were measured at the beginning and the end of the run. Results: There were no differences between groups in VALR, VILR, vertical stiffness, or tibial shock in a fresh or exerted condition. However, compared to N-BSI, women with M-BSI had greater increase with exertion in VALR (-1.8% vs. 6.1%, p = 0.01) and VILR (1.5% vs. 4.8%, p = 0.03). Similarly, compared to N-BSI, vertical stiffness increased more with exertion among women with M-BSI (-0.9% vs. 7.3%, p = 0.006) and 1-BSI (-0.9% vs. 1.8%, p = 0.05). Finally, compared to N-BSI, the increase in tibial shock from fresh to exerted condition was greater among women with M-BSI (0.9% vs. 5.5%, p = 0.03) and 1-BSI (0.9% vs. 11.2%, p = 0.02). Conclusion: Women with 1-BSI or M-BSIs experience greater exertion-related increases in impact loading than women with N-BSI. These observations imply that exertion-related changes in gait biomechanics may contribute to risk of BSI.

While some studies have found strong correlations between peak tibial accelerations (TAs) and early stance ground reaction forces (GRFs) during running, others have reported inconsistent results. One potential explanation for this is the lack of a standard orientation for the sensors used to collect TAs. Therefore, our aim was to test the effects of an established sensor reorientation method on peak Tas and their correlations with GRFs. Twenty-eight runners had TA and GRF data collected while they ran at a self-selected speed on an instrumented treadmill. Tibial accelerations were reoriented to a body-fixed frame using a simple calibration trial involving quiet standing and kicking. The results showed significant differences between raw and reoriented peak TAs (p < 0.01) for all directions except for the posterior (p = 0.48). The medial and lateral peaks were higher (+0.9–1.3 g), while the vertical and anterior were lower (−0.5–1.6 g) for reoriented vs. raw accelerations. Correlations with GRF measures were generally higher for reoriented TAs, although these differences were fairly small (Δr2 = 0.04–0.07) except for lateral peaks (Δr2 = 0.18). While contingent on the position of the IMU on the tibia used in our study, our results first showed systematic differences between reoriented and raw peak accelerations. However, we did not find major improvements in correlations with GRF measures for the reorientation method. This method may still hold promise for further investigation and development, given that consistent increases in correlations were found.

Inertial measurement units (IMUs) attached to the distal tibia are a validated method of measuring lower-extremity impact accelerations, called tibial accelerations (TAs), in runners. However, no studies have investigated the effects of small errors in IMU placement, which would be expected in real-world, autonomous use of IMUs. The purpose of this study was to evaluate the effect of a small proximal shift in IMU location on mean TAs and relationships between TAs and ground reaction force loading rates. IMUs were strapped to 18 injury-free runners at a specified standard location (∼1 cm proximal to medial malleolus) and 2 cm proximal to the standard location. TAs and ground reaction forces were measured while participants ran at self-selected and 10% slower/faster speeds. Mean TA was lower at the standard versus proximal IMU location in the faster running condition (P = .026), but similar in the slower (P = .643) and self-selected conditions (P = .654). Mean TAs measured at the standard IMU explained more variation in ground reaction force loading rates (r2 = .79-.90; P < .001) compared with those measured at the proximal IMU (r2 = .65-.72; P < .001). These results suggest that careful attention should be given to IMU placement when measuring TAs during running.

Background: The presence of bone marrow edema (BME) on magnetic resonance imaging (MRI) has been used to evaluate for bone stress injuries in athletes. Purpose: To examine the prevalence of MRI findings, including BME, in a single male collegiate basketball team before and after a single season and to assess its association with clinically symptomatic metatarsal bone stress injuries. Study Design: Cohort Study; Level of evidence, 3. Methods: A total of 16 men on a single collegiate basketball team (mean age, 20.0 ± 1.8 years) underwent 1.5-T MRI focused on both midfeet during the preseason, and 13 underwent repeat MRI during the postseason. MRI findings included the presence of BME and the radiographic classification of the bone stress injury (grades 1-4). Injury surveillance performed by athletic trainers was used to identify metatarsal bone stress injuries over the course of the season. Results: Preseason MRI demonstrated metatarsal BME in 5 of the 16 participants, and postseason MRI demonstrated metatarsal BME in 4 of the 13 participants. All 4 of the participants with postseason BME had MRI findings of BME in the same metatarsals. Compared to those without BME, participants with metatarsal BME had a shorter history of basketball exposure (preseason: 10.4 ± 4.1 vs 14.2 ± 1.9 years, respectively [P = .023]; postseason: 9.6 ± 4.1 vs 14.0 ± 2.1 years, respectively [P = .024]), and those with postseason BME had started playing at an older age (9.8 ± 4.3 vs 6.2 ± 1.6 years, respectively; P = .050). The preseason MRI classification for metatarsals included grade 1 (n = 3), followed by grades 2 and 3 (n = 2 each). In the 4 participants with postseason MRI findings, the grade increased from 1 to 4 in 1 participant and was stable in the other 3. No participants were diagnosed clinically with a metatarsal bone stress injury during the season. BME of the sesamoids was identified in 6 participants, who trended toward being older (21.0 ± 2.2 vs 19.4 ± 1.3 years, respectively; P < .10), with the abnormalities persisting on postseason MRI in all players. Conclusion: Collegiate male basketball players may have a high prevalence of BME, often without associated symptoms. The absence of foot pain or a corresponding diagnosis of a metatarsal bone stress injury in this study suggests that MRI findings of BME in asymptomatic athletes should be interpreted with caution.

Markerless motion capture allows whole-body movements to be captured without the need for physical markers to be placed on the body. This enables motion capture analyses to be conducted in more ecologically valid environments. However, the influences of varied clothing on video-based markerless motion capture assessments remain largely unexplored. This study investigated two types of clothing conditions, "Sport" (gym shirt and shorts) and "Street" (unrestricted casual clothing), on gait parameters during overground walking by 29 participants at self-selected speeds using markerless motion capture. Segment lengths, gait spatiotemporal parameters, and lower-limb kinematics were compared between the two clothing conditions. Mean differences in segment length for the forearm, upper arm, thigh, and shank between clothing conditions ranged from 0.2 cm for the forearm to 0.9 cm for the thigh (p < 0.05 for thigh and shank) but below typical marker placement errors (1 - 2 cm). Seven out of 9 gait spatiotemporal parameters demonstrated statistically significant differences between clothing conditions (p < 0.05), however, these differences were approximately ten times smaller than minimal detectable changes in movement-related pathologies including multiple sclerosis and cerebral palsy. Hip, knee, and ankle joint angle root-mean-square deviation values averaged 2.6° and were comparable to previously reported average inter-session variability for this markerless system (2.8°). The results indicate that clothing, a potential limiting factor in markerless motion capture performance, would negligibly alter meaningful clinical interpretations under the conditions investigated.

Several open-source platforms for markerless motion capture offer the ability to track 2-dimensional (2D) kinematics using simple digital video cameras. We sought to establish the performance of one of these platforms, DeepLabCut. Eighty-four runners who had sagittal plane videos recorded of their left lower leg were included in the study. Data from 50 participants were used to train a deep neural network for 2D pose estimation of the foot and tibia segments. The trained model was used to process novel videos from 34 participants for continuous 2D coordinate data. Overall network accuracy was assessed using the train/test errors. Foot and tibia angles were calculated for 7 strides using manual digitization and markerless methods. Agreement was assessed with mean absolute differences and intraclass correlation coefficients. Bland-Altman plots and paired t tests were used to assess systematic bias. The train/test errors for the trained network were 2.87/7.79 pixels, respectively (0.5/1.2 cm). Compared to manual digitization. the markerless method was found to systematically overestimate foot angles and underestimate tibial angles (P < .01, d = 0.06-0.26). However, excellent agreement was found between the segment calculation methods, with mean differences <= 1 degrees and intraclass correlation coefficients >=.90. Overall, these results demonstrate that open-source. markerless methods are a promising new tool for analyzing human motion.